Randomized Binary Consensus with Faulty Agents

This paper investigates self-organizing binary majority consensus disturbed by 1 faulty nodes with random and persistent failure. We study consensus in ordered and random 2 networks with noise, message loss, and delays. Using computer simulations, we show that: 3 (a) explicit randomization by noise, message loss, and topology can increase robustness 4 towards faulty nodes, (b) commonly-used faulty nodes with random failure inhibit consensus 5 less than faulty nodes with persistent failure, and (c) in some cases such randomly failing 6 faulty nodes can even promote agreement. 7